CN109087340A - A kind of face three-dimensional rebuilding method and system comprising dimensional information - Google Patents
A kind of face three-dimensional rebuilding method and system comprising dimensional information Download PDFInfo
- Publication number
- CN109087340A CN109087340A CN201810566350.2A CN201810566350A CN109087340A CN 109087340 A CN109087340 A CN 109087340A CN 201810566350 A CN201810566350 A CN 201810566350A CN 109087340 A CN109087340 A CN 109087340A
- Authority
- CN
- China
- Prior art keywords
- dimensional
- face
- point
- standard faces
- threedimensional model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/20—Finite element generation, e.g. wire-frame surface description, tesselation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/55—Depth or shape recovery from multiple images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
- G06T2207/30201—Face
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Processing Or Creating Images (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to technical field of computer vision, embodiment specifically discloses a kind of face three-dimensional rebuilding method and system comprising dimensional information.Face three-dimensional rebuilding method provided by the invention and system, it will be from standard faces face two dimensional character point obtained in standard faces threedimensional model, it is matched with the target person face portion two dimensional character point extracted from target face two dimensional image, standard faces threedimensional model is adjusted, initial target human face three-dimensional model is obtained；The three-dimensional feature point that will be extracted from target face three-dimensional point cloud is matched with the three-dimensional feature point extracted from initial target human face three-dimensional model, adjusts target face three-dimensional point cloud, obtains final goal face three-dimensional point cloud.It solves the problems, such as that existing method cannot rebuild the human face three-dimensional model with dimensional information and surface information, is not deformed and possessed the face three-dimensional reconstruction result of surface details, realize human face three-dimensional model broader practice.
Description
Technical field
The present invention relates to technical field of computer vision, and in particular to a kind of face three-dimensional reconstruction side comprising dimensional information
Method and system.
Background technique
Traditional three-dimensional rebuilding method has: photometric stereo, structure light, binocular vision, TOF etc..Photometric stereo vision can
To regard a kind of three-dimensional surface rebuilding method as, the body surface details obtained by it is very rich, and reduction degree is high, but traditional
Photometric stereo vision on the whole due to model inexactness and there are deviations, and can not directly be used there are dimensional variation
In three-dimensional measurement, reconstructing the threedimensional model come may deform.Structure light and TOF scheduling algorithm result are also more accurate,
But some demanding equipment are all relied on, application scenarios are than more limited.The reconstruction knot that binocular stereo vision obtains
Fruit has specific dimensional information, but noise is larger, and algorithm is not very robust.So according to mentioned-above three-dimensional reconstruction side
The all each have their own limitation of the reconstructed results that method obtains.
Currently, more and more fields need to carry out three-dimensional reconstruction to face, to obtain the ruler of face three-dimensional stereo model
Information and surface information are spent, therefore, how to obtain the face three-dimensional reconstruction result that one did not deform and possessed surface details is
Urgent problem to be solved at present.
Summary of the invention
In view of this, the application provides the scheme of a kind of combination human face characteristic point and face three-dimensional point cloud, can overcome on
Disadvantage is stated, the face three-dimensional point cloud with dimensional information and surface details information is obtained.
In order to solve the above technical problems, technical solution provided by the invention is a kind of face Three-dimensional Gravity comprising dimensional information
Construction method, comprising:
Obtain the target face two dimensional image of target face；
N number of two dimensional character point of target face two dimensional image is extracted from the target face two dimensional image；
Obtain standard faces threedimensional model；
N number of three-dimensional feature point of standard faces threedimensional model is extracted from the standard faces threedimensional model；
According to N number of three-dimensional feature point of the standard faces threedimensional model, N number of two dimension of standard faces threedimensional model is obtained
Characteristic point；
By N number of two dimension of N number of the two dimensional character point and the target face two dimensional image of the standard faces threedimensional model
Characteristic point is matched, and adjusts the standard faces threedimensional model, obtains initial target human face three-dimensional model；
N number of three-dimensional feature of initial target human face three-dimensional model is extracted from the initial target human face three-dimensional model
Point；
Obtain the target face three-dimensional point cloud of target face；
N number of three-dimensional feature point of target face three-dimensional point cloud is extracted from the target face three-dimensional point cloud；
By the N number of of N number of three-dimensional feature point of the target face three-dimensional point cloud and the initial target human face three-dimensional model
Three-dimensional feature point is matched, and adjusts the target face three-dimensional point cloud, obtains final goal face three-dimensional point cloud；
Wherein N is the positive integer greater than 1.
Preferably, N number of three-dimensional feature point according to the standard faces threedimensional model obtains standard faces three-dimensional mould
The method of N number of two dimensional character point of type, comprising: N number of three-dimensional feature point of the standard faces threedimensional model is projected into two dimension
Plane obtains N number of two dimensional character point of standard faces threedimensional model.
Preferably, N number of two dimensional character point by the standard faces threedimensional model and the target face X-Y scheme
N number of two dimensional character point of picture is matched, and adjusts the standard faces threedimensional model, obtains initial target human face three-dimensional model
Method, comprising:
Adjust N number of two dimensional character point and N number of the two of the target face two dimensional image of the standard faces threedimensional model
Dimensional feature point is overlapped, and obtains N number of two dimensional character point of standard faces threedimensional model adjusted；
N number of two dimensional character point back projection of the standard faces threedimensional model adjusted is three-dimensional to the standard faces
Model obtains N number of three-dimensional feature point of standard faces threedimensional model adjusted；
The standard faces three-dimensional mould is adjusted according to N number of three-dimensional feature point of the standard faces threedimensional model adjusted
Type obtains initial target human face three-dimensional model.
Preferably, described that the mark is adjusted according to N number of three-dimensional feature point of the standard faces threedimensional model adjusted
Quasi- human face three-dimensional model, the method for obtaining initial target human face three-dimensional model, comprising:
Construct the N number of three-dimensional feature point and the standard faces threedimensional model of the standard faces threedimensional model adjusted
N number of three-dimensional feature point characteristic point difference energy function；
To the N number of three-dimensional feature point and the standard faces threedimensional model of the standard faces threedimensional model adjusted
Each characteristic point difference energy value of N number of three-dimensional feature point is summed, and is carried out global solution by optimization algorithm, is obtained practical adjustment
The coordinate of N number of three-dimensional feature point of standard faces threedimensional model afterwards；
According to standard described in the Coordinate Adjusting of N number of three-dimensional feature point of reality standard faces threedimensional model adjusted
Human face three-dimensional model obtains initial target human face three-dimensional model.
Preferably, the method for the target face three-dimensional point cloud for obtaining target face, comprising: use photometric stereo vision
The target face three-dimensional point cloud of method acquisition target face.
Preferably, N number of three-dimensional feature point by the target face three-dimensional point cloud and the initial target face three
N number of three-dimensional feature point of dimension module is matched, and adjusts the target face three-dimensional point cloud, and it is three-dimensional to obtain final goal face
The method of point cloud, comprising:
Adjust the N of N number of the three-dimensional feature point and the initial target human face three-dimensional model of the target face three-dimensional point cloud
A three-dimensional feature point is overlapped, and obtains N number of three-dimensional feature point of target face three-dimensional point cloud adjusted；
The target face three-dimensional point is adjusted according to N number of three-dimensional feature point of the target face three-dimensional point cloud adjusted
Cloud obtains final goal face three-dimensional point cloud.
Preferably, described that the mesh is adjusted according to N number of three-dimensional feature point of the target face three-dimensional point cloud adjusted
Face three-dimensional point cloud is marked, the method for obtaining final goal face three-dimensional point cloud, comprising:
Construct the N number of three-dimensional feature point and the target face three-dimensional point cloud of the target face three-dimensional point cloud adjusted
N number of three-dimensional feature point characteristic point difference energy function；
To the N number of three-dimensional feature point and the target face three-dimensional point cloud of the target face three-dimensional point cloud adjusted
Each characteristic point difference energy value of N number of three-dimensional feature point is summed, and is carried out global solution by optimization algorithm, is obtained practical adjustment
The coordinate of N number of three-dimensional feature point of target face three-dimensional point cloud afterwards；
According to target described in the Coordinate Adjusting of N number of three-dimensional feature point of reality target face three-dimensional point cloud adjusted
Face three-dimensional point cloud obtains final goal face three-dimensional point cloud.
The present invention also provides a kind of face three-dimensional reconstruction system comprising dimensional information characterized by comprising
Target face two dimensional image obtains module, for obtaining the target face two dimensional image of target face；
Target face two dimensional image two dimensional character point extraction module, for extracting target from target face two dimensional image
N number of two dimensional character point of face two dimensional image；
Standard faces obtaining three-dimensional model module, for obtaining standard faces threedimensional model；
Standard faces threedimensional model three-dimensional feature point extraction module, for extracting standard from standard faces threedimensional model
N number of three-dimensional feature point of human face three-dimensional model；
Standard faces threedimensional model two dimensional character point obtains module, for N number of three-dimensional according to standard faces threedimensional model
Characteristic point obtains N number of two dimensional character point of standard faces threedimensional model；
Two dimensional character point Matching and modification module, for by the N number of two dimensional character point and target person of standard faces threedimensional model
N number of two dimensional character point of face two dimensional image is matched, and adjusts standard faces threedimensional model, and it is three-dimensional to obtain initial target face
Model；
Initial target human face three-dimensional model three-dimensional feature point extraction module, for being mentioned from initial target human face three-dimensional model
Take out N number of three-dimensional feature point of initial target human face three-dimensional model；
Target face three-dimensional point cloud obtains module, for obtaining the target face three-dimensional point cloud of target face；
Target face three-dimensional point cloud three-dimensional feature point extraction module, for extracting target from target face three-dimensional point cloud
N number of three-dimensional feature point of face three-dimensional point cloud；
Three-dimensional feature point Matching and modification module, for by N number of three-dimensional feature point of target face three-dimensional point cloud and initial mesh
N number of three-dimensional feature point of mark human face three-dimensional model is matched, and adjusts target face three-dimensional point cloud, obtains final goal face
Three-dimensional point cloud.
Preferably, the two dimensional character point Matching and modification module, comprising:
First matching unit, for adjusting the N number of two dimensional character point and target face X-Y scheme of standard faces threedimensional model
N number of two dimensional character point of picture is overlapped, and obtains N number of two dimensional character point of standard faces threedimensional model adjusted；
First projecting cell, for by N number of two dimensional character point back projection of standard faces threedimensional model adjusted to mark
Quasi- human face three-dimensional model obtains N number of three-dimensional feature point of standard faces threedimensional model adjusted；
The first adjustment unit, for adjusting standard according to N number of three-dimensional feature point of standard faces threedimensional model adjusted
Human face three-dimensional model obtains initial target human face three-dimensional model.
Preferably, the three-dimensional feature point Matching and modification module, comprising:
Second matching unit, for adjusting the N number of three-dimensional feature point and initial target face three of target face three-dimensional point cloud
N number of three-dimensional feature point of dimension module is overlapped, and obtains N number of three-dimensional feature point of target face three-dimensional point cloud adjusted；
Second adjustment unit, for adjusting target according to N number of three-dimensional feature point of target face three-dimensional point cloud adjusted
Face three-dimensional point cloud obtains final goal face three-dimensional point cloud.
Compared with prior art, detailed description are as follows for its advantages by the application: provided in an embodiment of the present invention includes ruler
The face three-dimensional rebuilding method and system of information are spent, it will be special from the face two dimension of standard faces obtained in standard faces threedimensional model
Point is levied, is matched with the target person face portion two dimensional character point extracted from target face two dimensional image, standard faces are adjusted
Threedimensional model, to obtain initial target human face three-dimensional model；The three-dimensional feature point that will be extracted from target face three-dimensional point cloud,
It is matched with the three-dimensional feature point extracted from initial target human face three-dimensional model, adjusts target face three-dimensional point cloud, thus
Obtain final goal face three-dimensional point cloud.Solving existing three-dimensional rebuilding method cannot rebuild with dimensional information and surface letter
The problem of human face three-dimensional model of breath, obtains the face three-dimensional reconstruction result for not deforming and possessing surface details, real
Human face three-dimensional model broader practice is showed.
Detailed description of the invention
Fig. 1 is the method flow schematic diagram of face three-dimensional reconstruction of the embodiment of the present invention one comprising dimensional information；
Fig. 2 is the method flow schematic diagram that the embodiment of the present invention one obtains initial target human face three-dimensional model；
Fig. 3 is that the embodiment of the present invention one is marked according to the adjustment of N number of three-dimensional feature point of standard faces threedimensional model adjusted
The method flow schematic diagram of quasi- human face three-dimensional model；
Fig. 4 is the method flow schematic diagram that the embodiment of the present invention one obtains final goal face three-dimensional point cloud；
Fig. 5 is the embodiment of the present invention one according to N number of three-dimensional feature point of target face three-dimensional point cloud adjusted adjustment mesh
Mark the method flow schematic diagram of face three-dimensional point cloud；
Fig. 6 is the flow diagram of face three-dimensional rebuilding method of the embodiment of the present invention two comprising dimensional information；
Fig. 7 is the structural schematic diagram of face three-dimensional reconstruction system of the embodiment of the present invention three comprising dimensional information；
Fig. 8 is the structural schematic diagram of face three-dimensional reconstruction system of the embodiment of the present invention four comprising dimensional information.
Specific embodiment
It is with reference to the accompanying drawing and specific real in order to make those skilled in the art more fully understand technical solution of the present invention
Applying example, the present invention is described in further detail.
As shown in Figure 1, the embodiment of the invention provides a kind of face three-dimensional rebuilding method comprising dimensional information, this method
Include:
S1: the target face two dimensional image of target face is obtained；
S2: N number of two dimensional character point of target face two dimensional image is extracted from target face two dimensional image；
S3: standard faces threedimensional model is obtained；
S4: N number of three-dimensional feature point of standard faces threedimensional model is extracted from standard faces threedimensional model；
S5: according to N number of three-dimensional feature point of standard faces threedimensional model, N number of two dimension of standard faces threedimensional model is obtained
Characteristic point；
S6: N number of two dimension of N number of two dimensional character point of standard faces threedimensional model and the target face two dimensional image is special
Sign point is matched, and adjusts standard faces threedimensional model, obtains initial target human face three-dimensional model；
S7: N number of three-dimensional feature point of initial target human face three-dimensional model is extracted from initial target human face three-dimensional model；
S8: the target face three-dimensional point cloud of target face is obtained；
S9: N number of three-dimensional feature point of target face three-dimensional point cloud is extracted from target face three-dimensional point cloud；
S10: by N number of three-dimensional of N number of the three-dimensional feature point and initial target human face three-dimensional model of target face three-dimensional point cloud
Characteristic point is matched, and adjusts target face three-dimensional point cloud, obtains final goal face three-dimensional point cloud；
Wherein N is the positive integer greater than 1.
It should be noted that in step S1, the method for obtaining the target face two dimensional image of target face can pass through phase
Machine photographic subjects face two dimensional image, can also directly receive target face two dimensional image.Here in order to guarantee target face three
Tie up the effect rebuild, the target face two dimensional image of acquisition, it is desirable that can completely extract the target person mentioned in step S2
N number of two dimensional character point of face two dimensional image.
The target face two dimensional image of acquisition can be the target face front two dimensional image of target person face, the target face
Front two dimensional image requires that N number of human face characteristic point can be extracted, and may include eyebrow, eyes, nose, mouth and profile et al.
Face characteristic point.The target face two dimensional image of acquisition tries not to block, if the target face two dimensional image obtained occurs
It blocks or certain human face characteristic points is extracted less than can have an impact to subsequent reconstructed results.It is therefore desirable to the target obtained
Face two dimensional image does not block, and can extract N number of face two dimensional character point.
The target face two dimensional image of acquisition can be black white image, infrared image or color image.Target face two
The higher the better for the pixel of dimension image, can obtain the better face three-dimensional reconstruction result of effect.Color image with infrared image or
The difference of person's black white image is that color image can carry out texture mapping for reconstructing the human face three-dimensional model come, obtains
The human face rebuilding result of coloured multimedia message.If it is black white image or infrared image is used, then without this step of texture mapping
Suddenly.
In step S2, N number of two dimensional character point of target face two dimensional image is extracted from target face two dimensional image
Method can use ASM algorithm or deep learning algorithm.
ASM algorithm is a kind of method of the feature point extraction based on Statistical learning model.It is not popularized in deep learning
Before, ASM should be that human face characteristic point extracts preferable solution.ASM algorithm is similar with common statistical learning algorithm, needs
First to prepare the data set of tape label, Train training (or Build or foundation) model, then Test tests (Fit matching).
Deep learning algorithm is the common method that current human face characteristic point extracts, according to be trained the model that draws come
Extract characteristic point.Selected characteristic point is all manually to be marked on many facial images for training early period, then will
It is trained inside the tag image input deep learning network marked, obtains model parameter.Finally when input one open it is new
When unmarked image, the training parameter that it can use front directly finds out characteristic point.
The embodiment extracts the N of target face two dimensional image using deep learning algorithm from target face two dimensional image
A two dimensional character point.
Here the number N of the two dimensional character point extracted is to need selection according to our application.Such as when we conduct oneself
When face is aligned, we may select five or six characteristic points of eyes nose mouth.Here we are for extracting target
The model of N number of two dimensional character point of face two dimensional image can use the existing model of other databases.Existing face characteristic
Point, which extracts model, to be had including more than 30 a human face characteristic points, also has including 68 human face characteristic points, also has including 78 faces
Characteristic point, there are also include 83 human face characteristic points.
It is adjusted after needing to match by human face characteristic point due to us in the embodiment, it is possible to select human face characteristic point
The a little more model of number.In order to obtain the better face three-dimensional modeling of effect as a result, the embodiment can be using comprising 68
The human face characteristic point of a human face characteristic point extracts model, and target face two dimensional image is extracted from target face two dimensional image
68 two dimensional character points.
In step S3, the standard faces threedimensional model of acquisition can be some data that the current research field has had
The face standard three-dimensional model in library is also possible to the face standard three-dimensional model that we are obtained ourselves by deep learning.
In step S4, N number of three-dimensional feature point of standard faces threedimensional model is extracted from standard faces threedimensional model
Method can use deep learning algorithm.Deep learning algorithm is the common method that current human face characteristic point extracts, according to progress
The model extraction characteristic point that training is drawn.Selected characteristic point is all manually to be used for above trained faceforms many early period
It is marked, then will be trained inside the markup model marked input deep learning network, obtain model parameter.Finally
When one new unmarked model of input, the training parameter that it can use front directly finds out characteristic point.
The embodiment extracts the N of standard faces threedimensional model using deep learning algorithm from standard faces threedimensional model
A three-dimensional feature point.
Here the number N of the three-dimensional feature point extracted is to need selection according to our application.In the embodiment due to
We adjust after needing to match by human face characteristic point, it is possible to the model for selecting human face characteristic point number a little more.For
The better face three-dimensional modeling of effect is obtained as a result, the embodiment can use the face characteristic comprising 68 human face characteristic points
Point extracts model, and 68 three-dimensional feature points of standard faces threedimensional model are extracted from standard faces threedimensional model.
In step S5, according to N number of three-dimensional feature point of standard faces threedimensional model, the N of standard faces threedimensional model is obtained
The method of a two dimensional character point, can be with are as follows: N number of three-dimensional feature point of standard faces threedimensional model is projected to two-dimensional surface, is obtained
Obtain N number of two dimensional character point of standard faces threedimensional model.
Here, standard people can also be converted to by other methods by N number of three-dimensional feature point of standard faces threedimensional model
N number of two dimensional character point of face three-dimensional model, as long as the method that can be realized same effect is ok.
As shown in Fig. 2, in step S6, by N number of two dimensional character point of standard faces threedimensional model and target face X-Y scheme
N number of two dimensional character point of picture is matched, and adjusts standard faces threedimensional model, obtains the side of initial target human face three-dimensional model
Method, can be with are as follows:
S61: N number of two dimensional character point of adjustment standard faces threedimensional model and the two dimensional character point of target face two dimensional image
It is overlapped, obtains N number of two dimensional character point of standard faces threedimensional model adjusted.
Here, by N number of two dimensional character of N number of the two dimensional character point and target face two dimensional image of standard faces threedimensional model
Point carries out matched method, can be to be and target person by the Coordinate Adjusting of N number of two dimensional character point of standard faces threedimensional model
The coordinate of N number of two dimensional character point of face two dimensional image is overlapped.
S62: by N number of two dimensional character point back projection of standard faces threedimensional model adjusted to standard faces three-dimensional mould
Type obtains N number of three-dimensional feature point of standard faces threedimensional model adjusted.
Here, by N number of two dimensional character point back projection of standard faces threedimensional model adjusted to standard faces three-dimensional mould
Type, the method for obtaining N number of three-dimensional feature point of standard faces threedimensional model adjusted can be with are as follows: the standard faces three of acquisition
The coordinate of one three-dimensional feature point A of dimension module is (x1, y1, z1), and A point is projected to two-dimensional surface, obtains corresponding two
Dimensional feature point A ' (x1, y1) becomes tune after corresponding points a ' (x2, the y2) alignment in point A ' and target face two dimensional image
A ' (x2, y2) after whole, by A ' adjusted (x2, y2) back projection to standard faces threedimensional model, obtain A adjusted (x2,
Y2, z2), wherein x2, y2 value is unchanged, and the value of z2 is adjusted according to the variation relation of the xy of points several around A, can use down
The energy function in face carries out energy and minimizes the z2 coordinate for solving to the end:
Wherein N (i) is some three-dimensional feature point i
Adjoint point.
S63: adjusting standard faces threedimensional model according to N number of three-dimensional feature point of standard faces threedimensional model adjusted,
Obtain initial target human face three-dimensional model.
As shown in figure 3, it is three-dimensional to adjust standard faces according to N number of three-dimensional feature point of standard faces threedimensional model adjusted
Model, the method for obtaining initial target human face three-dimensional model, can be with are as follows:
S631: the N number of three-dimensional feature point for constructing standard faces threedimensional model adjusted and the standard faces three before adjustment
The characteristic point difference energy function of N number of three-dimensional feature point of dimension module.
S632: the standard faces before N number of three-dimensional feature point and adjustment to standard faces threedimensional model adjusted are three-dimensional
Each characteristic point difference energy value of N number of three-dimensional feature point of model is summed, and is carried out global solution by optimization algorithm, is obtained real
The coordinate of N number of three-dimensional feature point of border standard faces threedimensional model adjusted.
S633: according to the Coordinate Adjusting standard people of N number of three-dimensional feature point of practical standard faces threedimensional model adjusted
Face three-dimensional model obtains initial target human face three-dimensional model.
It should be noted that the energy function that construction is total are as follows:
Wherein Ii(x, y, z) indicates adjusted
The coordinate of N number of three-dimensional feature point of standard faces threedimensional model,Standard faces three before indicating adjustment
The coordinate of N number of three-dimensional feature point of dimension module.Global solution is carried out by optimization algorithm, changes total energy value most
It is small, the coordinate of N number of three-dimensional feature point of practical standard faces threedimensional model adjusted is finally solved, after practical adjustment
Standard faces threedimensional model N number of three-dimensional feature point Coordinate Adjusting standard faces threedimensional model, obtain initial target face
Threedimensional model.
In step S7, N number of three-dimensional of initial target human face three-dimensional model is extracted from initial target human face three-dimensional model
The method of characteristic point can use deep learning algorithm.Deep learning algorithm is the common method that current human face characteristic point extracts,
According to being trained the model extraction characteristic point drawn.Selected characteristic point is all early period manually in many faces for training
It is marked above model, then will be trained inside the markup model marked input deep learning network, obtain model
Parameter.Finally when one new unmarked model of input, the training parameter that it can use front directly finds out characteristic point.
The embodiment extracts initial target face three using deep learning algorithm from initial target human face three-dimensional model
N number of three-dimensional feature point of dimension module.
Here the number N of the three-dimensional feature point extracted is to need selection according to our application.In the embodiment due to
We adjust after needing to match by human face characteristic point, it is possible to the model for selecting human face characteristic point number a little more.For
The better face three-dimensional modeling of effect is obtained as a result, the embodiment can use the face characteristic comprising 68 human face characteristic points
Point extracts model, and 68 three-dimensional features of initial target human face three-dimensional model are extracted from initial target human face three-dimensional model
Point.
In step S8, the method for obtaining the target face three-dimensional point cloud of target face can use photometric stereo vision side
The target face three-dimensional point cloud of method acquisition target face.
Photometric stereo visible sensation method can regard a kind of three-dimensional surface rebuilding method as, and it is non-to obtain face surface details by it
Often abundant, reduction degree is high.And compared to other methods, photometric stereo algorithmic technique has the advantage that 1, equipment requirement is simple, easily
In operation；2, the reconstruction of pixel scale, point cloud are dense；3, the three-dimensional face reconstructed has good details surface characteristics, letter
Breath reduction is accurate.Therefore, photometric stereo algorithm has important theoretical significance and practical application value.
Here, the method for obtaining the target face three-dimensional point cloud of target face has very much, but photometric stereo visible sensation method
It is the current progress higher algorithm of three-dimensional surface rebuilding precision, the target person of target face is obtained using photometric stereo visible sensation method
Face three-dimensional point cloud can reconstruct the better human face three-dimensional model of effect.
Here, the method that the target face three-dimensional point cloud of target face is obtained using photometric stereo visible sensation method, can wrap
It includes: being irradiated in turn using camera and a infrared LED lamp light source of n (n > 2), obtain the facial image under n different angle illumination, benefit
With luminosity Stereo Vision, the face three-dimensional point cloud with fine surface is obtained.Wherein, photometric stereo visible sensation method can
To include: 1, Source calibration step；2, normal estimation step；3, curve reestablishing step.
It should be noted that the target face three-dimensional point cloud obtained using other methods, the perhaps essence of three-dimensional surface rebuilding
Degree is not highest, as long as but can satisfy the demands of application scenarios, can similarly use.As long as the mesh of requirement can be obtained
The method for marking face three-dimensional point cloud, can use.
In step S9, N number of three-dimensional feature point of target face three-dimensional point cloud is extracted from target face three-dimensional point cloud
Method, can be using the method for the N number of three-dimensional feature point for extracting target face three-dimensional point cloud according to geometrical characteristic.It can also adopt
With deep learning algorithm.Deep learning algorithm is the common method that current human face characteristic point extracts, and is drawn according to being trained
Model extraction characteristic point.Selected characteristic point is all manually to be marked on many faceforms for training early period,
Then it will be trained inside the markup model marked input deep learning network, obtain model parameter.Finally when input one
A new unmarked model, the training parameter that it can use front directly find out characteristic point.
The embodiment extracts the N of target face three-dimensional point cloud using deep learning algorithm from target face three-dimensional point cloud
A three-dimensional feature point.
Here the number N of the three-dimensional feature point extracted is to need selection according to our application.In the embodiment due to
We need to be registrated adjustment later by human face characteristic point, it is possible to the model for selecting human face characteristic point number a little more.
In order to obtain the better face three-dimensional modeling of effect as a result, the embodiment can be special using the face comprising 68 human face characteristic points
Sign point extracts model, and 68 three-dimensional feature points of target face three-dimensional point cloud are extracted from target face three-dimensional point cloud.
As shown in figure 4, in step S10, by N number of three-dimensional feature point of target face three-dimensional point cloud and initial target face three
N number of three-dimensional feature point of dimension module is matched, and adjusts target face three-dimensional point cloud, obtains final goal face three-dimensional point cloud
Method, can be with are as follows:
S101: N number of three-dimensional feature point of adjustment target face three-dimensional point cloud and N number of the three of initial target human face three-dimensional model
Dimensional feature point is overlapped, and obtains N number of three-dimensional feature point of target face three-dimensional point cloud adjusted.
Here by N number of three-dimensional of N number of the three-dimensional feature point and initial target human face three-dimensional model of target face three-dimensional point cloud
Characteristic point is matched is and initial target face as by the Coordinate Adjusting of N number of three-dimensional feature point of target face three-dimensional point cloud
The coordinate of N number of three-dimensional feature point of threedimensional model is overlapped.
S102: adjusting target face three-dimensional point cloud according to N number of three-dimensional feature point of target face three-dimensional point cloud adjusted,
Obtain final goal face three-dimensional point cloud.
As shown in figure 5, it is three-dimensional to adjust target face according to N number of three-dimensional feature point of target face three-dimensional point cloud adjusted
Point cloud, the method for obtaining final goal face three-dimensional point cloud can be with are as follows:
S1021: the N number of three-dimensional feature point and the target face three before adjustment for constructing target face three-dimensional point cloud adjusted
The characteristic point difference energy function of N number of three-dimensional feature point of dimension point cloud.
S1022: the target face before N number of three-dimensional feature point and adjustment to target face three-dimensional point cloud adjusted is three-dimensional
Each characteristic point difference energy value summation of N number of three-dimensional feature point of point cloud carries out global solution by optimization algorithm, obtains real
The coordinate of N number of three-dimensional feature point of border target face three-dimensional point cloud adjusted.
S1023: according to the Coordinate Adjusting target person of N number of three-dimensional feature point of practical target face three-dimensional point cloud adjusted
Face three-dimensional point cloud obtains final goal face three-dimensional point cloud.
It should be noted that the energy function that construction is total are as follows:Its
Middle Ii(x, y, z) indicates the coordinate of N number of three-dimensional feature point of target face three-dimensional point cloud adjusted,
The coordinate of N number of three-dimensional feature point of target face three-dimensional point cloud before indicating adjustment.Global solution is carried out by optimization algorithm,
Change total energy value minimum, finally solves N number of three-dimensional feature point of practical target face three-dimensional point cloud adjusted
Coordinate, the Coordinate Adjusting target face according to N number of three-dimensional feature point of practical target face three-dimensional point cloud adjusted is three-dimensional
Point cloud, obtains final goal face three-dimensional point cloud.
As shown in fig. 6, the embodiment of the present invention two also provides another face three-dimensional rebuilding method comprising dimensional information, it should
Method includes:
S10: the target face two dimensional image of target face is obtained；
S20: N number of two dimensional character point of target face two dimensional image is extracted from target face two dimensional image；
S30: standard faces threedimensional model is obtained；
S40: N number of three-dimensional feature point of standard faces threedimensional model is extracted from standard faces threedimensional model；
S50: according to N number of three-dimensional feature point of standard faces threedimensional model, N number of two dimension of standard faces threedimensional model is obtained
Characteristic point；
S60: by N number of two dimension of N number of the two dimensional character point and the target face two dimensional image of standard faces threedimensional model
Characteristic point is matched, and adjusts standard faces threedimensional model, obtains initial target human face three-dimensional model；
S70: N number of three-dimensional feature of initial target human face three-dimensional model is extracted from initial target human face three-dimensional model
Point；
S80: the target face three-dimensional point cloud of target face is obtained；
S90: N number of three-dimensional feature point of target face three-dimensional point cloud is extracted from target face three-dimensional point cloud；
S100: by N number of three-dimensional of N number of the three-dimensional feature point and initial target human face three-dimensional model of target face three-dimensional point cloud
Characteristic point is matched, and adjusts target face three-dimensional point cloud, obtains final goal face three-dimensional point cloud；
S110: texture mapping is carried out to final goal face three-dimensional point cloud, obtains the final goal people with color information
Face three-dimensional point cloud；
Wherein N is the positive integer greater than 1.
It should be noted that the method for step S10-S100 is the same as example 1, this embodiment increases step S110,
In step S110, texture mapping is carried out to final goal face three-dimensional point cloud, obtains the final goal face with color information
Three-dimensional point cloud.Here target face two dimensional image can be obtained by the target face Two-dimensional Color Image of acquisition target face
Texture information, then to final goal face three-dimensional point cloud carry out texture mapping, obtain have color information final goal people
Face three-dimensional point cloud.
As shown in fig. 7, the embodiment of the present invention three also provides a kind of face three-dimensional reconstruction system comprising dimensional information, packet
It includes:
Target face two dimensional image obtains module, for obtaining the target face two dimensional image of target face.
Target face two dimensional image two dimensional character point extraction module, for extracting target from target face two dimensional image
N number of two dimensional character point of face two dimensional image.
Standard faces obtaining three-dimensional model module, for obtaining standard faces threedimensional model.
Standard faces threedimensional model three-dimensional feature point extraction module, for extracting standard from standard faces threedimensional model
N number of three-dimensional feature point of human face three-dimensional model.
Standard faces threedimensional model two dimensional character point obtains module, for N number of three-dimensional according to standard faces threedimensional model
Characteristic point obtains N number of two dimensional character point of standard faces threedimensional model.
Two dimensional character point Matching and modification module, for by the N number of two dimensional character point and target person of standard faces threedimensional model
N number of two dimensional character point of face two dimensional image is matched, and adjusts standard faces threedimensional model, and it is three-dimensional to obtain initial target face
Model.
Initial target human face three-dimensional model three-dimensional feature point extraction module, for being mentioned from initial target human face three-dimensional model
Take out N number of three-dimensional feature point of initial target human face three-dimensional model.
Target face three-dimensional point cloud obtains module, for obtaining the target face three-dimensional point cloud of target face.
Target face three-dimensional point cloud three-dimensional feature point extraction module, for extracting target from target face three-dimensional point cloud
N number of three-dimensional feature point of face three-dimensional point cloud.
Three-dimensional feature point Matching and modification module, for by N number of three-dimensional feature point of target face three-dimensional point cloud and initial mesh
N number of three-dimensional feature point of mark human face three-dimensional model is matched, and adjusts target face three-dimensional point cloud, obtains final goal face
Three-dimensional point cloud.
Wherein, two dimensional character point Matching and modification module, comprising:
First matching unit, for adjusting the N number of two dimensional character point and target face X-Y scheme of standard faces threedimensional model
The two dimensional character point of picture is overlapped, and obtains N number of two dimensional character point of standard faces threedimensional model adjusted.
First projecting cell, for by N number of two dimensional character point back projection of standard faces threedimensional model adjusted to mark
Quasi- human face three-dimensional model obtains N number of three-dimensional feature point of standard faces threedimensional model adjusted；
The first adjustment unit, for adjusting standard according to N number of three-dimensional feature point of standard faces threedimensional model adjusted
Human face three-dimensional model obtains initial target human face three-dimensional model.
The first adjustment unit includes:
First construction component, for constructing N number of three-dimensional feature point and the standard people of standard faces threedimensional model adjusted
The characteristic point difference energy function of N number of three-dimensional feature point of face three-dimensional model.
First computation module, for the N number of three-dimensional feature point and standard faces to standard faces threedimensional model adjusted
Each characteristic point difference energy value of N number of three-dimensional feature point of threedimensional model is summed, and is carried out global solution by optimization algorithm, is obtained
Obtain the coordinate of N number of three-dimensional feature point of practical standard faces threedimensional model adjusted.
The first adjustment component, for the seat according to actually N number of three-dimensional feature point of standard faces threedimensional model adjusted
Mark adjustment standard faces threedimensional model, obtains initial target human face three-dimensional model.
It should be noted that three-dimensional feature point Matching and modification module, comprising:
Second matching unit, for adjusting the N number of three-dimensional feature point and initial target face three of target face three-dimensional point cloud
N number of three-dimensional feature point of dimension module is overlapped, and obtains N number of three-dimensional feature point of target face three-dimensional point cloud adjusted.
Second adjustment unit, for adjusting target according to N number of three-dimensional feature point of target face three-dimensional point cloud adjusted
Face three-dimensional point cloud obtains final goal face three-dimensional point cloud.
Wherein, second adjustment unit includes:
Second construction component, for constructing the N number of three-dimensional feature point and target person of target face three-dimensional point cloud adjusted
The characteristic point difference energy function of N number of three-dimensional feature point of face three-dimensional point cloud.
Second computing module, for the N number of three-dimensional feature point and target face to target face three-dimensional point cloud adjusted
Each characteristic point difference energy value of N number of three-dimensional feature point of three-dimensional point cloud is summed, and is carried out global solution by optimization algorithm, is obtained
Obtain the coordinate of N number of three-dimensional feature point of practical target face three-dimensional point cloud adjusted.
Second adjustment module, for the seat according to actually N number of three-dimensional feature point of target face three-dimensional point cloud adjusted
Mark adjustment target face three-dimensional point cloud, obtains final goal face three-dimensional point cloud.
As shown in figure 8, the embodiment of the present invention four also provides another face three-dimensional reconstruction system comprising dimensional information, packet
It includes:
Target face two dimensional image obtains module, for obtaining the target face two dimensional image of target face.
Target face two dimensional image two dimensional character point extraction module, for extracting target from target face two dimensional image
N number of two dimensional character point of face two dimensional image.
Standard faces obtaining three-dimensional model module, for obtaining standard faces threedimensional model.
Standard faces threedimensional model three-dimensional feature point extraction module, for extracting standard from standard faces threedimensional model
N number of three-dimensional feature point of human face three-dimensional model.
Standard faces threedimensional model two dimensional character point obtains module, for N number of three-dimensional according to standard faces threedimensional model
Characteristic point obtains N number of two dimensional character point of standard faces threedimensional model.
Two dimensional character point Matching and modification module, for by the N number of two dimensional character point and target person of standard faces threedimensional model
N number of two dimensional character point of face two dimensional image is matched, and adjusts standard faces threedimensional model, and it is three-dimensional to obtain initial target face
Model.
Initial target human face three-dimensional model three-dimensional feature point extraction module, for being mentioned from initial target human face three-dimensional model
Take out N number of three-dimensional feature point of initial target human face three-dimensional model.
Target face three-dimensional point cloud obtains module, for obtaining the target face three-dimensional point cloud of target face.
Target face three-dimensional point cloud three-dimensional feature point extraction module, for extracting target from target face three-dimensional point cloud
N number of three-dimensional feature point of face three-dimensional point cloud.
Three-dimensional feature point Matching and modification module, for by N number of three-dimensional feature point of target face three-dimensional point cloud and initial mesh
N number of three-dimensional feature point of mark human face three-dimensional model is matched, and adjusts target face three-dimensional point cloud, obtains final goal face
Three-dimensional point cloud.
Texture mapping module, for carrying out texture mapping to final goal face three-dimensional point cloud, obtaining has color information
Final goal face three-dimensional point cloud.
It should be noted that example IV on the basis of embodiment three, increases texture mapping module.The texture mapping
Module by the target face Two-dimensional Color Image of acquisition target face, can obtain the texture letter of target face two dimensional image
Breath, then texture mapping is carried out to final goal face three-dimensional point cloud, obtain the final goal face three-dimensional point with color information
Cloud.
The above is only the preferred embodiment of the present invention, it is noted that above-mentioned preferred embodiment is not construed as pair
Limitation of the invention, protection scope of the present invention should be defined by the scope defined by the claims..For the art
For those of ordinary skill, without departing from the spirit and scope of the present invention, several improvements and modifications can also be made, these change
It also should be regarded as protection scope of the present invention into retouching.
Claims (10)
1. a kind of face three-dimensional rebuilding method comprising dimensional information characterized by comprising
Obtain the target face two dimensional image of target face；
N number of two dimensional character point of target face two dimensional image is extracted from the target face two dimensional image；
Obtain standard faces threedimensional model；
N number of three-dimensional feature point of standard faces threedimensional model is extracted from the standard faces threedimensional model；
According to N number of three-dimensional feature point of the standard faces threedimensional model, N number of two dimensional character of standard faces threedimensional model is obtained
Point；
By N number of two dimensional character of N number of the two dimensional character point and the target face two dimensional image of the standard faces threedimensional model
Point is matched, and adjusts the standard faces threedimensional model, obtains initial target human face three-dimensional model；
N number of three-dimensional feature point of initial target human face three-dimensional model is extracted from the initial target human face three-dimensional model；
Obtain the target face three-dimensional point cloud of target face；
N number of three-dimensional feature point of target face three-dimensional point cloud is extracted from the target face three-dimensional point cloud；
By N number of three-dimensional of N number of the three-dimensional feature point and the initial target human face three-dimensional model of the target face three-dimensional point cloud
Characteristic point is matched, and adjusts the target face three-dimensional point cloud, obtains final goal face three-dimensional point cloud；
Wherein N is the positive integer greater than 1.
2. the face three-dimensional rebuilding method according to claim 1 comprising dimensional information, which is characterized in that described according to institute
N number of three-dimensional feature point of standard faces threedimensional model is stated, the method for obtaining N number of two dimensional character point of standard faces threedimensional model,
Include: that N number of three-dimensional feature point of the standard faces threedimensional model is projected into two-dimensional surface, obtains standard faces threedimensional model
N number of two dimensional character point.
3. the face three-dimensional rebuilding method according to claim 1 comprising dimensional information, which is characterized in that it is described will be described
N number of two dimensional character point of standard faces threedimensional model is matched with N number of two dimensional character point of the target face two dimensional image,
And the standard faces threedimensional model is adjusted, the method for obtaining initial target human face three-dimensional model, comprising:
N number of two dimension of the N number of two dimensional character point and the target face two dimensional image that adjust the standard faces threedimensional model is special
Sign point is overlapped, and obtains N number of two dimensional character point of standard faces threedimensional model adjusted；
By N number of two dimensional character point back projection of the standard faces threedimensional model adjusted to the standard faces three-dimensional mould
Type obtains N number of three-dimensional feature point of standard faces threedimensional model adjusted；
The standard faces threedimensional model is adjusted according to N number of three-dimensional feature point of the standard faces threedimensional model adjusted,
Obtain initial target human face three-dimensional model.
4. the face three-dimensional rebuilding method according to claim 3 comprising dimensional information, which is characterized in that described according to institute
The N number of three-dimensional feature point for stating standard faces threedimensional model adjusted adjusts the standard faces threedimensional model, obtains initial mesh
The method for marking human face three-dimensional model, comprising:
Construct the N of N number of the three-dimensional feature point and the standard faces threedimensional model of the standard faces threedimensional model adjusted
The characteristic point difference energy function of a three-dimensional feature point；
To the N number of of N number of three-dimensional feature point of the standard faces threedimensional model adjusted and the standard faces threedimensional model
Each characteristic point difference energy value of three-dimensional feature point is summed, and global solution is carried out by optimization algorithm, after obtaining practical adjustment
Standard faces threedimensional model N number of three-dimensional feature point coordinate；
According to standard faces described in the Coordinate Adjusting of N number of three-dimensional feature point of reality standard faces threedimensional model adjusted
Threedimensional model obtains initial target human face three-dimensional model.
5. the face three-dimensional rebuilding method according to claim 1 comprising dimensional information, which is characterized in that the acquisition mesh
The method for marking the target face three-dimensional point cloud of face, comprising: the target person of target face is obtained using photometric stereo visible sensation method
Face three-dimensional point cloud.
6. the face three-dimensional rebuilding method according to claim 1 comprising dimensional information, which is characterized in that it is described will be described
N number of three-dimensional feature point of target face three-dimensional point cloud and N number of three-dimensional feature point of the initial target human face three-dimensional model carry out
Matching, and the target face three-dimensional point cloud is adjusted, the method for obtaining final goal face three-dimensional point cloud, comprising:
Adjust N number of three-dimensional feature point and N number of the three of the initial target human face three-dimensional model of the target face three-dimensional point cloud
Dimensional feature point is overlapped, and obtains N number of three-dimensional feature point of target face three-dimensional point cloud adjusted；
The target face three-dimensional point cloud is adjusted according to N number of three-dimensional feature point of the target face three-dimensional point cloud adjusted,
Obtain final goal face three-dimensional point cloud.
7. the face three-dimensional rebuilding method according to claim 6 comprising dimensional information, which is characterized in that described according to institute
The N number of three-dimensional feature point for stating target face three-dimensional point cloud adjusted adjusts the target face three-dimensional point cloud, obtains final mesh
The method for marking face three-dimensional point cloud, comprising:
Construct the N of N number of the three-dimensional feature point and the target face three-dimensional point cloud of the target face three-dimensional point cloud adjusted
The characteristic point difference energy function of a three-dimensional feature point；
To the N number of of N number of three-dimensional feature point of the target face three-dimensional point cloud adjusted and the target face three-dimensional point cloud
Each characteristic point difference energy value of three-dimensional feature point is summed, and global solution is carried out by optimization algorithm, after obtaining practical adjustment
Target face three-dimensional point cloud N number of three-dimensional feature point coordinate；
According to target face described in the Coordinate Adjusting of N number of three-dimensional feature point of reality target face three-dimensional point cloud adjusted
Three-dimensional point cloud obtains final goal face three-dimensional point cloud.
8. a kind of face three-dimensional reconstruction system comprising dimensional information characterized by comprising
Target face two dimensional image obtains module, for obtaining the target face two dimensional image of target face；
Target face two dimensional image two dimensional character point extraction module, for extracting target face from target face two dimensional image
N number of two dimensional character point of two dimensional image；
Standard faces obtaining three-dimensional model module, for obtaining standard faces threedimensional model；
Standard faces threedimensional model three-dimensional feature point extraction module, for extracting standard faces from standard faces threedimensional model
N number of three-dimensional feature point of threedimensional model；
Standard faces threedimensional model two dimensional character point obtains module, for N number of three-dimensional feature according to standard faces threedimensional model
Point obtains N number of two dimensional character point of standard faces threedimensional model；
Two dimensional character point Matching and modification module, for by N number of two dimensional character point of standard faces threedimensional model and target face two
N number of two dimensional character point of dimension image is matched, and adjusts standard faces threedimensional model, obtains initial target face three-dimensional mould
Type；
Initial target human face three-dimensional model three-dimensional feature point extraction module, for being extracted from initial target human face three-dimensional model
N number of three-dimensional feature point of initial target human face three-dimensional model；
Target face three-dimensional point cloud obtains module, for obtaining the target face three-dimensional point cloud of target face；
Target face three-dimensional point cloud three-dimensional feature point extraction module, for extracting target face from target face three-dimensional point cloud
N number of three-dimensional feature point of three-dimensional point cloud；
Three-dimensional feature point Matching and modification module, for by N number of three-dimensional feature point of target face three-dimensional point cloud and initial target people
N number of three-dimensional feature point of face three-dimensional model is matched, and adjusts target face three-dimensional point cloud, and it is three-dimensional to obtain final goal face
Point cloud.
9. the face three-dimensional reconstruction system according to claim 8 comprising dimensional information, which is characterized in that the two dimension is special
Sign point Matching and modification module, comprising:
First matching unit, the N of N number of two dimensional character point and target face two dimensional image for adjusting standard faces threedimensional model
A two dimensional character point is overlapped, and obtains N number of two dimensional character point of standard faces threedimensional model adjusted；
First projecting cell, for by N number of two dimensional character point back projection of standard faces threedimensional model adjusted to standard people
Face three-dimensional model obtains N number of three-dimensional feature point of standard faces threedimensional model adjusted；
The first adjustment unit, for adjusting standard faces according to N number of three-dimensional feature point of standard faces threedimensional model adjusted
Threedimensional model obtains initial target human face three-dimensional model.
10. the face three-dimensional reconstruction system according to claim 8 comprising dimensional information, which is characterized in that the three-dimensional
Feature Points Matching adjusts module, comprising:
Second matching unit, for adjusting the N number of three-dimensional feature point and initial target face three-dimensional mould of target face three-dimensional point cloud
N number of three-dimensional feature point of type is overlapped, and obtains N number of three-dimensional feature point of target face three-dimensional point cloud adjusted；
Second adjustment unit, for adjusting target face according to N number of three-dimensional feature point of target face three-dimensional point cloud adjusted
Three-dimensional point cloud obtains final goal face three-dimensional point cloud.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810566350.2A CN109087340A (en) | 2018-06-04 | 2018-06-04 | A kind of face three-dimensional rebuilding method and system comprising dimensional information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810566350.2A CN109087340A (en) | 2018-06-04 | 2018-06-04 | A kind of face three-dimensional rebuilding method and system comprising dimensional information |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109087340A true CN109087340A (en) | 2018-12-25 |
Family
ID=64839316
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810566350.2A Pending CN109087340A (en) | 2018-06-04 | 2018-06-04 | A kind of face three-dimensional rebuilding method and system comprising dimensional information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109087340A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110263617A (en) * | 2019-04-30 | 2019-09-20 | 北京永航科技有限公司 | Three-dimensional face model acquisition methods and device |
| CN111639553A (en) * | 2020-05-14 | 2020-09-08 | 青岛联合创智科技有限公司 | Preparation method of customized mask device based on visual three-dimensional reconstruction |
| WO2021004257A1 (en) * | 2019-07-10 | 2021-01-14 | 广州市百果园信息技术有限公司 | Line-of-sight detection method and apparatus, video processing method and apparatus, and device and storage medium |
| CN112347870A (en) * | 2020-10-23 | 2021-02-09 | 歌尔光学科技有限公司 | Image processing method, device and equipment of head-mounted equipment and storage medium |
| CN112396692A (en) * | 2020-11-25 | 2021-02-23 | 北京市商汤科技开发有限公司 | Face reconstruction method and device, computer equipment and storage medium |
| WO2021249401A1 (en) * | 2020-06-08 | 2021-12-16 | 北京京东乾石科技有限公司 | Model generation method and apparatus, image perspective determining method and apparatus, device, and medium |
| CN114998527A (en) * | 2022-06-27 | 2022-09-02 | 上海域圆信息科技有限公司 | High-accuracy three-dimensional human body surface reconstruction system |
| RU2782543C1 (en) * | 2019-07-10 | 2022-10-31 | Биго Текнолоджи Пте. Лтд. | Method and device for sight line detection, method and device for video data processing, device and data carrier |
| WO2023272725A1 (en) * | 2021-07-02 | 2023-01-05 | 华为技术有限公司 | Facial image processing method and apparatus, and vehicle |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1920886A (en) * | 2006-09-14 | 2007-02-28 | 浙江大学 | Video flow based three-dimensional dynamic human face expression model construction method |
| CN101101672A (en) * | 2007-07-13 | 2008-01-09 | 中国科学技术大学 | Stereo vision three-dimensional human face modelling approach based on dummy image |
| CN101303772A (en) * | 2008-06-20 | 2008-11-12 | 浙江大学 | Method for modeling non-linear three-dimensional human face based on single sheet image |
| CN101339669A (en) * | 2008-07-29 | 2009-01-07 | 上海师范大学 | Three-dimensional human face modelling approach based on front side image |
| CN101593365A (en) * | 2009-06-19 | 2009-12-02 | 电子科技大学 | A kind of method of adjustment of universal three-dimensional human face model |
| CN101625768A (en) * | 2009-07-23 | 2010-01-13 | 东南大学 | Three-dimensional human face reconstruction method based on stereoscopic vision |
| CN101916454A (en) * | 2010-04-08 | 2010-12-15 | 董洪伟 | Method for reconstructing high-resolution human face based on grid deformation and continuous optimization |
| CN103606190A (en) * | 2013-12-06 | 2014-02-26 | 上海明穆电子科技有限公司 | Method for automatically converting single face front photo into three-dimensional (3D) face model |
| CN104680582A (en) * | 2015-03-24 | 2015-06-03 | 中国人民解放军国防科学技术大学 | Method for creating object-oriented customized three-dimensional human body model |
| CN107154074A (en) * | 2017-01-18 | 2017-09-12 | 南京悦容网络科技有限公司 | A kind of three-dimensional face modeling method of feature based |
| US20170365092A1 (en) * | 2016-06-21 | 2017-12-21 | Apple Inc. | Method and System for Vision Based 3D Reconstruction and Object Tracking |
| CN107862732A (en) * | 2017-11-08 | 2018-03-30 | 清华大学 | Real-time three-dimensional eyelid method for reconstructing and device |
-
2018
- 2018-06-04 CN CN201810566350.2A patent/CN109087340A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1920886A (en) * | 2006-09-14 | 2007-02-28 | 浙江大学 | Video flow based three-dimensional dynamic human face expression model construction method |
| CN101101672A (en) * | 2007-07-13 | 2008-01-09 | 中国科学技术大学 | Stereo vision three-dimensional human face modelling approach based on dummy image |
| CN101303772A (en) * | 2008-06-20 | 2008-11-12 | 浙江大学 | Method for modeling non-linear three-dimensional human face based on single sheet image |
| CN101339669A (en) * | 2008-07-29 | 2009-01-07 | 上海师范大学 | Three-dimensional human face modelling approach based on front side image |
| CN101593365A (en) * | 2009-06-19 | 2009-12-02 | 电子科技大学 | A kind of method of adjustment of universal three-dimensional human face model |
| CN101625768A (en) * | 2009-07-23 | 2010-01-13 | 东南大学 | Three-dimensional human face reconstruction method based on stereoscopic vision |
| CN101916454A (en) * | 2010-04-08 | 2010-12-15 | 董洪伟 | Method for reconstructing high-resolution human face based on grid deformation and continuous optimization |
| CN103606190A (en) * | 2013-12-06 | 2014-02-26 | 上海明穆电子科技有限公司 | Method for automatically converting single face front photo into three-dimensional (3D) face model |
| CN104680582A (en) * | 2015-03-24 | 2015-06-03 | 中国人民解放军国防科学技术大学 | Method for creating object-oriented customized three-dimensional human body model |
| US20170365092A1 (en) * | 2016-06-21 | 2017-12-21 | Apple Inc. | Method and System for Vision Based 3D Reconstruction and Object Tracking |
| CN107154074A (en) * | 2017-01-18 | 2017-09-12 | 南京悦容网络科技有限公司 | A kind of three-dimensional face modeling method of feature based |
| CN107862732A (en) * | 2017-11-08 | 2018-03-30 | 清华大学 | Real-time three-dimensional eyelid method for reconstructing and device |
Non-Patent Citations (6)
| Title |
|---|
| HUI CHEN等: "Feature points matching for face reconstruction based on the window unique property of pseudo-random coded image", 《MECHATRONICS》 * |
| 林道贵等: "基于真实人脸的数字化三维人脸建模", 《福建教育学院学报》 * |
| 涂意等: "基于单张人脸图片和一般模型的三维重建方法", 《计算机应用研究》 * |
| 王剑等: "图像三维重建中的特征点匹配方法研究", 《计算机工程与应用》 * |
| 程日彬等: "基于二维图像的三维人脸建模技术", 《计算机工程与应用》 * |
| 黄明伟等: "改进泊松算法的图像三维重建点云模型网格化", 《测绘科学》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110263617A (en) * | 2019-04-30 | 2019-09-20 | 北京永航科技有限公司 | Three-dimensional face model acquisition methods and device |
| CN110263617B (en) * | 2019-04-30 | 2021-10-22 | 北京永航科技有限公司 | Three-dimensional face model obtaining method and device |
| WO2021004257A1 (en) * | 2019-07-10 | 2021-01-14 | 广州市百果园信息技术有限公司 | Line-of-sight detection method and apparatus, video processing method and apparatus, and device and storage medium |
| US12039752B2 (en) | 2019-07-10 | 2024-07-16 | Bigo Technology Pte. Ltd. | Method for determining line-of-sight, method for processing video, device, and storage medium |
| RU2782543C1 (en) * | 2019-07-10 | 2022-10-31 | Биго Текнолоджи Пте. Лтд. | Method and device for sight line detection, method and device for video data processing, device and data carrier |
| CN111639553A (en) * | 2020-05-14 | 2020-09-08 | 青岛联合创智科技有限公司 | Preparation method of customized mask device based on visual three-dimensional reconstruction |
| WO2021249401A1 (en) * | 2020-06-08 | 2021-12-16 | 北京京东乾石科技有限公司 | Model generation method and apparatus, image perspective determining method and apparatus, device, and medium |
| CN112347870B (en) * | 2020-10-23 | 2023-03-24 | 歌尔科技有限公司 | Image processing method, device and equipment of head-mounted equipment and storage medium |
| CN112347870A (en) * | 2020-10-23 | 2021-02-09 | 歌尔光学科技有限公司 | Image processing method, device and equipment of head-mounted equipment and storage medium |
| WO2022110855A1 (en) * | 2020-11-25 | 2022-06-02 | 北京市商汤科技开发有限公司 | Face reconstruction method and apparatus, computer device, and storage medium |
| CN112396692A (en) * | 2020-11-25 | 2021-02-23 | 北京市商汤科技开发有限公司 | Face reconstruction method and device, computer equipment and storage medium |
| CN112396692B (en) * | 2020-11-25 | 2023-11-28 | 北京市商汤科技开发有限公司 | Face reconstruction method, device, computer equipment and storage medium |
| WO2023272725A1 (en) * | 2021-07-02 | 2023-01-05 | 华为技术有限公司 | Facial image processing method and apparatus, and vehicle |
| CN114998527A (en) * | 2022-06-27 | 2022-09-02 | 上海域圆信息科技有限公司 | High-accuracy three-dimensional human body surface reconstruction system |
| CN114998527B (en) * | 2022-06-27 | 2024-05-03 | 上海域圆信息科技有限公司 | High-accuracy three-dimensional human body surface reconstruction system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109087340A (en) | A kind of face three-dimensional rebuilding method and system comprising dimensional information | |
| CN105631861B (en) | Restore the method for 3 D human body posture from unmarked monocular image in conjunction with height map | |
| CN107204010B (en) | A kind of monocular image depth estimation method and system | |
| CN111784821B (en) | Three-dimensional model generation method and device, computer equipment and storage medium | |
| CN108960020A (en) | Information processing method and information processing equipment | |
| CN109035388A (en) | Three-dimensional face model method for reconstructing and device | |
| CN104794722A (en) | Dressed human body three-dimensional bare body model calculation method through single Kinect | |
| CN104299250B (en) | Front face image synthetic method and system based on prior model | |
| CN101968892A (en) | Method for automatically adjusting three-dimensional face model according to one face picture | |
| US8743119B2 (en) | Model-based face image super-resolution | |
| CN108121950B (en) | Large-pose face alignment method and system based on 3D model | |
| KR20170019779A (en) | Method and Apparatus for detection of 3D Face Model Using Portable Camera | |
| WO2024032464A1 (en) | Three-dimensional face reconstruction method, apparatus, and device, medium, and product | |
| CN105427385A (en) | High-fidelity face three-dimensional reconstruction method based on multilevel deformation model | |
| CN101968891A (en) | System for automatically generating three-dimensional figure of picture for game | |
| CN109086798A (en) | A kind of data mask method and annotation equipment | |
| CN110399809A (en) | The face critical point detection method and device of multiple features fusion | |
| CN112734890B (en) | Face replacement method and device based on three-dimensional reconstruction | |
| CN109903377A (en) | A kind of three-dimensional face modeling method and system without phase unwrapping | |
| WO2012078006A2 (en) | Image processor, lighting processor and method therefor | |
| TWI750710B (en) | Image processing method and apparatus, image processing device and storage medium | |
| CN109242834A (en) | It is a kind of based on convolutional neural networks without reference stereo image quality evaluation method | |
| CN109410119A (en) | Mask image distortion method and its system | |
| CN102567716A (en) | Face synthetic system and implementation method | |
| CN109146934A (en) | A kind of face three-dimensional rebuilding method and system based on binocular solid and photometric stereo |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181225 |